Tech news in 3 minutes
OpenAI limits GPT-5.6 rollout after government request, says restrictions shouldn’t be the norm
OpenAI announced Friday that it is limiting the release of its new GPT-5.6 lineup—comprising Sol (flagship), Terra (balanced), and Luna (fast, low-cost)—to a “small group of trusted partners” at the request of the U.S. government under the Trump administration. The government’s action follows increased pressure on AI companies to restrict advanced systems, exemplified by the administration ordering Anthropic to remove public access to its Fable 5 model after release, which Anthropic subsequently took down entirely. Dean Ball, a former White House AI adviser and incoming OpenAI employee, argued that President Trump’s recent executive order—asking AI companies to voluntarily submit advanced models for government review up to 30 days before release—has created a de facto involuntary licensing regime. Without clear safety standards, he warned, launch delays could cede ground to China in the AI race and threaten billions in AI infrastructure investment. OpenAI complied but expressed discontent, stating in a blog post that “this kind of government access process should not become the long-term default” as it keeps tools from users and global partners. The company called the preview a “short-term step” toward broader availability in coming weeks, while working with the administration on a new executive order framework for cybersecurity and a repeatable release process. GPT-5.6 Sol is OpenAI’s strongest model, with improved agentic capabilities in coding, biology, and cybersecurity, including a “max” reasoning mode and “ultra” mode using coordinated subagents. It benchmarks slightly better than Anthropic’s Claude Mythos 5 while using a third of the output tokens. OpenAI emphasized Sol’s robust security stack, hardened against adversarial attacks and optimized for defensive cybersecurity, with safety guardrails built into the core model to avoid the false-positive issues that plagued Anthropic’s Fable 5. The models will be made broadly available via ChatGPT, Codex, and API soon. Pricing: Sol at $5 input/$30 output per million tokens; Terra half that; Luna at $1/$6. OpenAI also improved prompt caching for repeated prompts.